While diabetes mellitus is recognized clinically as a complication of alcohol dependence, in the last 15 years several large prospective studies have demonstrated a significant reduction in risk of incident type 2 diabetes in moderate drinkers. In this article, we review prospective studies on the association between alcohol consumption and incidence of diabetes. Few prospective studies have demonstrated an unequivocal positive association between alcohol use and incidence of diabetes. A number of large prospective studies have now demonstrated an inverse association: typically a 40% risk reduction after controlling for other major risk factors. Studies based on cohorts of health professionals have not demonstrated an increased risk of diabetes with heavier consumption, but these cohorts contain few heavy drinkers. Some cohorts drawn from the general population have shown a J- or U-shaped association between level of alcohol consumption and incidence of diabetes. Frequent moderate drinking has been reported to be associated with lower risk than infrequent drinking. There have been contradictory reports on the importance of the type of alcoholic beverage and whether the effect of alcohol differs according to the bodyweight of the drinker. We conclude that like many other chronic diseases, there is a delicate balance between the harmful and beneficial effects of alcohol on the incidence of diabetes. In moderate amounts, drinking is associated with a reduced risk of diabetes, whereas in higher amounts with an increased risk.